Integrated Intelligent Prosthetic Control System: A reliability-aware, predictive and adaptive architecture for powered lower-limb prostheses
Year: 2026
Type: preprint
Abstract: Powered lower-limb prostheses increasingly rely on machine learning to infer wearer intent, yet in many existing systems the classifier output is treated directly as the motor command. This paper presents a complete conceptual system architecture in which authority over movement is deliberately allocated across five cooperating subsystems: wearer sensing, environmental perception, a supervisory unit, a local prosthesis controller, and the communication and power infrastructure. The architecture is predictive (anticipation supplies the lead time that authorisation requires), adaptive (wearer-specific patterns are learned within inviolable limits), and reliability-aware (evidence quality and subsystem health continuously shape permitted actions). The core contribution is the system itself, defined by six bounded, independently testable properties that must hold under stated assumptions even when individual subsystems are wrong, late, degraded or compromised: Authority, Validity, Local Independence, Degradation, Reconciliation and Wearer Intervention. The central principle is that a prediction is only evidence; it must earn permission from the limb's actual physical state before it becomes movement. Four distinct objects are kept separate throughout: what the system predicts, what it requests, what the prosthesis authorises, and what the prosthesis ultimately executes. This is a fully specified conceptual architecture. It has not yet been implemented or experimentally validated. Any claim of novelty or effectiveness requires a controller-level literature review, systematic hazard analysis (STPA), formal verification (model checking), and comparative evaluation (fault-injection on a test rig). The architecture is offered as a rigorous, falsifiable foundation for that programme of work.
Author: Yadav, Kartik
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225618545
Cites:
Cited by:
Related to: